AI/ML●●Solid
Parlor Jarvis – Realtime AI (audio+screen in, voice out) & multilingual
Multilingual local AI with screen sharing beats Parlor's English-only camera input.
Niche GemShip It
unusual_typo
821mo ago
Back to browse
GitHub Repository
On-device, real-time multimodal AI. Have natural voice and vision conversations with an AI that runs entirely on your machine. Powered by Gemma 4 E2B and Kokoro.
1,809 starsHTML
Real-time AI (audio/video in, voice out) on an M3 Pro with Gemma E2B
by karimf·Apr 5, 2026·298 points·38 comments
AI Analysis
●●●BangerWizardryDark Horse
Runs Gemma 4 E2B and Kokoro TTS locally with barge-in and vision.
Strengths
- •On-device processing eliminates server costs and privacy concerns for voice AI.
- •Barge-in and sentence-level streaming create natural conversation flow without pauses.
- •Combines vision and voice in a single local stack for interaction.
Weaknesses
- •Requires M3 Pro or strong GPU, limiting accessibility for older hardware.
- •Research preview means bugs and rough edges are expected in builds.
Category
Target Audience
Developers building voice AI, privacy-focused users, language learners
Similar To
OpenAI Realtime API · LiveKit Agents · Bule AI
Similar Projects
AI/ML●●●Banger
Gemma 4 Multimodal Fine-Tuner for Apple Silicon
Only Apple Silicon toolkit streaming GCS data during audio fine-tuning without OOM.
WizardryNiche GemZero to One
MediaSquirrel
235281mo ago
AI/ML●●●Banger
Qiaohu – offline multimodal voice assistant on Snapdragon 8 Gen 2
Full voice assistant pipeline with barge-in running entirely offline on Snapdragon GPU.
WizardryNiche GemZero to One
donge
201mo ago
AI/ML●Mid
I benchmarked Gemma 4 E2B – the 2B model beat the 12B on multi-turn
2B model beats 12B on some tasks, saving hardware costs for edge deployment.
Big BrainNiche Gem
mailharishin
811mo ago
AI/ML●●Solid
Running Gemma 4 on an iPhone 13 Pro
Clean Swift wrapper for Gemma 4 with vision and audio on iPhone.
Niche GemShip It
dengjiuhong
101mo ago
AI/ML●●●Banger
Prompt-to-Excalidraw demo with Gemma 4 E2B in the browser (3.1GB)
50-token compact code output beats raw 5,000-token Excalidraw JSON — clever compression.
WizardryBig Brain
teamchong
163621mo ago